[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: PUPA bugfixes
From: |
Marco Gerards |
Subject: |
Re: PUPA bugfixes |
Date: |
07 Nov 2003 21:24:35 +0100 |
User-agent: |
Gnus/5.09 (Gnus v5.9.0) Emacs/21.2 |
"Yoshinori K. Okuji" <address@hidden> writes:
> On Saturday 01 November 2003 17:41, Marco Gerards wrote:
> > Okuji, you told me it is better to check if a disk was swapped and if
> > not the disk cache should be flushed. I have implemented that, but I
> > don't really like this solution. Some BIOSes have bugs that prevent
> > this check from working and often it is impossible to check if the
> > BIOS has such bug. I also dislike that the check takes some time.
>
> What bug do you mean? Do you have a pointer to such information?
> Also, how much time does it take? 1ms, 100ms, 1s, or more?
A few hundred ms.
> > Now I have implemented this timeout globally, if there is a timeout
> > the entire cache is flushed for _all_ drives. Is that ok for you or
> > should I implement it so it works per disk?
>
> I'm not sure. I need more time to consider this issue.
>
> In my experience, disk caching is important only for floppies. Hard disks are
> not so important, because they have internal caching and they are quite fast.
> I don't know how fast CDROM emulation mode is.
>
> Normally, the user would normally use a single disk to boot up his/her
> machine. I guess the number of disks concerning with boot stages would be at
> most 3. So it might not be meaningful to manage disk caches in per disk
> basis.
>
> What do you think?
It is much easier to manage the cache globally and I see no real
disadvantages.
> > And what do you think about adding "pupa-emu"? I already sent in a
> > patch, but I didn't get feedback about it yet. Do you want it? If
> > you want it, what should I change before the patch can be committed?
>
> Feel free to commit it. I have no objection.
I'm making some improvements again. I will do this soon. :)
> BTW, your patch does not contain some files, such as times.h. It would be
> better if you use the option -N to diff.
Oops.
Thanks,
Marco